dsh-intelhub

Social-media posts, bookmarked pages, and half-formed notes scatter across a dozen places. When you need them back, keyword search misses and the original wording is gone, turning information into noise. dsh-intelhub solves exactly that: it turns the continuous inflow of information into a searchable, filterable, and exportable local knowledge base.

Ingestion is source-agnostic. A watched folder picks up files the moment a collection script writes them, a web page is archived in one command, or you jot down a judgment as plain text. Search combines semantic and keyword matching, then layers scalar filters for author, likes, stage, and tags, so querying how to configure a timeout hits the original post that says abort after 30 seconds of no response. Results carry author, like count, and file-block citations. When a research sprint ends, export to plain Markdown in one step and let Obsidian index it with zero plugins.

Everything runs locally. The first import downloads a roughly 30 MB embedding model once; after that there is no API key, no daemon, and documents never leave your machine. It fits researchers, writers, and content curators who ingest a lot of information daily, prefer a local Markdown or Obsidian workflow, and do not want to maintain a heavyweight database.